claude-auto 0.12.4

package/.claude-auto/validators/infra-commit-format.md

@@ -0,0 +1,37 @@
+---
+name: infra-commit-format
+description: Validates commit message format for config-only commits
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** This validator ONLY applies when ALL changed files are infrastructure/config files.
+
+**Config files include:**
+- package.json, package-lock.json, pnpm-lock.yaml
+- tsconfig.json, tsconfig.*.json
+- *.config.ts, *.config.js (vite.config.ts, vitest.config.ts, etc.)
+- wrangler.toml
+- .gitignore, .npmignore, .eslintrc*, .prettierrc*
+- ketchup-plan.md
+- Any file in .claude/ directory
+
+**ACK immediately if:**
+- ANY .ts or .tsx file (not *.config.ts) is in the changed files list
+- The commit includes behavioral code, not just config
+
+**When ONLY config files are changed:**
+- ACK if commit message starts with `chore(scope):` or `chore:` format
+- NACK if commit message uses `feat:`, `fix:`, `test:` for config-only changes
+
+**Examples:**
+- Good: `chore(deps): update vitest to 1.0`
+- Good: `chore: add ketchup-plan.md`
+- Bad: `feat: update package.json` (for config-only commit)
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/ketchup-plan-format.md

@@ -0,0 +1,42 @@
+---
+name: ketchup-plan-format
+description: Validates ketchup-plan.md structure and format
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Validate ketchup-plan.md when it is included in the commit.
+
+**ACK immediately if:**
+- ketchup-plan.md is not in the changed files
+
+**When ketchup-plan.md is changed, validate:**
+
+**Required structure:**
+- Must have `## TODO` section
+- Must have `## DONE` section
+
+**Burst format:**
+- Bursts should include `[depends: ...]` notation
+- Format: `- [ ] Burst N: description [depends: none]` or `[depends: N, M]`
+- Completed bursts in DONE should show commit hash: `- [x] Burst N: description (abc1234)`
+
+**Bottle naming:**
+- Bottles should be named by capability, not sequence number
+- Format: `### Bottle: SettingsMerger` not `### Bottle 1`
+
+**NACK if:**
+- ketchup-plan.md lacks TODO or DONE sections
+- Bursts are missing dependency notation
+- Bottles are named by number instead of capability
+
+**ACK if:**
+- ketchup-plan.md follows the required structure
+- Or ketchup-plan.md is not in the diff
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/new-code-requires-tests.md

@@ -0,0 +1,36 @@
+---
+name: new-code-requires-tests
+description: Enforces that new behavioral code has accompanying tests
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Validate .ts and .tsx files (not config files) in the diff.
+
+Let tests drive all code. New behavioral code requires tests.
+
+**NACK if:**
+- New .ts/.tsx files are added without corresponding .test.ts/.test.tsx files
+- New conditional branches (`if`, `else`, `? :`, `??`, `||`) appear without test coverage
+- Functions are added without any test calling them
+
+**ACK if:**
+- New .ts files have accompanying .test.ts files in the commit
+- The diff only modifies existing code (not adding new files)
+- The diff only contains config files (*.config.ts, tsconfig.json, etc.)
+- The diff only contains test files
+- The new code is in an `index.ts` barrel file (re-exports only)
+
+**Config files that don't need tests:**
+- package.json, tsconfig.json, wrangler.toml
+- vite.config.ts, vitest.config.ts
+- .gitignore, ketchup-plan.md
+
+**First behavioral .ts file rule:** The first non-config .ts file in a project must have a test.
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/no-comments.md

@@ -0,0 +1,35 @@
+---
+name: no-comments
+description: Enforces self-documenting code without inline comments
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Only validate .ts and .tsx files in the diff (not .test.ts, not .md, not config files).
+
+Code should be self-documenting. No comments to explain what code does.
+
+**NACK if the diff adds:**
+- Single-line comments: `// explanation`
+- Multi-line comments: `/* explanation */`
+- TODO comments: `// TODO:`, `// FIXME:`, `// HACK:`
+- Explanatory comments describing what the code does
+
+**ACK if:**
+- No inline comments are added in .ts/.tsx source files
+- The diff only contains .md, .json, .test.ts, or config files
+- JSDoc comments for exported public APIs (these document the contract, not implementation)
+- Comments that are part of string literals or template strings
+- License headers at the top of files
+
+**Refactoring guidance:** If code needs a comment to explain it, the code should be refactored instead:
+- Rename variables to be self-describing
+- Extract functions with descriptive names
+- Simplify complex logic
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/no-dangerous-git.md

@@ -0,0 +1,35 @@
+---
+name: no-dangerous-git
+description: Blocks dangerous git commands
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Validate git commands for dangerous operations.
+
+These operations are forbidden and cannot be overridden:
+
+**ALWAYS NACK if the command contains:**
+- `--force` or `-f` with push (destroys remote history)
+- `--no-verify` (bypasses pre-commit hooks)
+- `git reset --hard` on shared branches
+- `git push` to main/master with `--force`
+- `git commit --amend` on already-pushed commits
+
+**NACK reasoning:**
+- `--force` can destroy team members' work
+- `--no-verify` bypasses all safety checks
+- These are explicitly listed as "DO NOT EVER ALLOW" in CLAUDE.md
+
+**ACK if:**
+- The git command does not contain dangerous flags
+- The command is a normal commit, push, or other safe operation
+
+**No plea system override:** These rules cannot be bypassed with a plea. They require explicit human approval outside the commit validator system.
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/tcr-workflow.md

@@ -0,0 +1,31 @@
+---
+name: tcr-workflow
+description: Enforces Test && Commit || Revert workflow
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Validate commit context for TCR workflow compliance.
+
+Enforce the TCR (Test && Commit || Revert) workflow:
+
+**Core principle:** Red → Green → TCR → Refactor → TCR → Done
+
+**NACK if:**
+- The commit message or context indicates tests are failing
+- The commit appears to be "patching" failing code rather than reverting and rethinking
+- Evidence suggests the commit bypasses the test-first workflow
+
+**ACK if:**
+- The commit appears to follow the TCR pattern
+- Tests are expected to pass (behavioral code has accompanying tests)
+- The commit represents a clean, tested change
+
+**Note:** This validator cannot run tests itself. It validates the commit context and message for TCR compliance indicators.
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/testing-no-state-peeking.md

@@ -0,0 +1,37 @@
+---
+name: testing-no-state-peeking
+description: Prohibits testing internal state instead of behavior
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Only validate .test.ts and .test.tsx files in the diff.
+
+Tests must verify observable behavior, not internal state.
+
+**NACK if the diff contains:**
+- Private field access: `obj["privateField"]`, `service["internalMap"]`
+- Bracket notation to access non-public properties
+- Calls to methods that exist solely for testing: `.getCount()`, `.getActiveSessionCount()`, `.getInternalState()`
+- Direct assertions on internal collections: `expect(tracker.items.length)`
+
+**Litmus test:** "If I changed the internal data structure (e.g., Map to Array), would this test still pass?" If no, it's testing implementation.
+
+**ACK if:**
+- Tests verify via return values from public methods
+- Tests verify via callbacks/events
+- Tests verify via thrown exceptions
+- Tests use the public API only
+- The diff only contains non-test files
+
+**Allowed patterns:**
+- `tracker.onEvent(e => events.push(e))` - observable via callback
+- `expect(createUser({name: 'Alice'})).toEqual({...})` - observable via return value
+- Testing cleanup by re-triggering and verifying callback fires again
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/testing-structure.md

@@ -0,0 +1,37 @@
+---
+name: testing-structure
+description: Enforces test structure and whole object assertions
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Only validate .test.ts and .test.tsx files in the diff.
+
+Enforce proper test structure:
+
+**NACK if the diff contains:**
+- Multiple execute/verify cycles in one test (squint test violation)
+- Cherry-picked property assertions instead of whole object assertions
+  - Bad: `expect(result.id).toBe(1); expect(result.name).toBe('foo');`
+  - Good: `expect(result).toEqual({ id: 1, name: 'foo' })`
+- Tests with multiple `expect()` calls on different operations
+
+**Required structure (Squint Test):**
+1. SETUP - prepare test data
+2. EXECUTE - single function call
+3. VERIFY - whole object assertion
+
+**ACK if:**
+- Tests follow SETUP → EXECUTE → VERIFY pattern
+- Whole objects are asserted with `toEqual({...})`
+- One assertion per behavior
+- The diff only contains non-test files
+
+**Note:** Multiple `expect()` calls are OK if they assert different aspects of the SAME operation result (e.g., checking both return value and side effect callback).
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/testing-stubs-over-mocks.md

@@ -0,0 +1,42 @@
+---
+name: testing-stubs-over-mocks
+description: Prefers deterministic stubs over mocks
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Only validate .test.ts and .test.tsx files in the diff.
+
+Prefer deterministic stubs over mocks. Mock only at system boundaries.
+
+**NACK if the diff contains:**
+- Heavy mocking of internal modules/functions that could use stubs
+- `jest.mock()` or `vi.mock()` for internal application code
+- Mock implementations that could be replaced with simple stub objects
+- Spy assertions on internal function calls (`toHaveBeenCalledWith` on non-boundary code)
+
+**ACK if:**
+- Deterministic stubs are used (plain objects/functions with predictable behavior)
+- Mocks are only used at system boundaries:
+  - External APIs (fetch, axios calls)
+  - Databases
+  - File system
+  - Third-party services
+- The diff only contains non-test files
+
+**Stub example (preferred):**
+```typescript
+const fakeRepo = { findById: (id) => ({ id, name: 'Test' }) }
+```
+
+**Mock example (only at boundaries):**
+```typescript
+vi.mock('./external-api', () => ({ fetch: vi.fn() }))
+```
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/testing-weak-assertions.md

@@ -0,0 +1,36 @@
+---
+name: testing-weak-assertions
+description: Prohibits weak test assertions
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Only validate .test.ts and .test.tsx files in the diff.
+
+Enforce strong assertions that verify exact values:
+
+**NACK if the diff contains:**
+- `toBeDefined()` - assert the actual value instead
+- `toBeTruthy()` - assert the exact truthy value
+- `toBeFalsy()` - assert the exact falsy value
+- `not.toBeNull()` - assert what it actually is
+- `not.toBeUndefined()` - assert the actual value
+- `toBeGreaterThan(0)` when exact value is known
+
+**Preferred alternatives:**
+- `toEqual({...})` for objects
+- `toBe(exactValue)` for primitives
+- `expect.any(String)` or `expect.any(Number)` when exact value is unknown but type is known
+- `toMatch(/pattern/)` for string patterns
+
+**ACK if:**
+- Tests use strong assertions (`toEqual`, `toBe`, `toMatch`, `toThrow`)
+- `expect.any()` is used for dynamic values like IDs or timestamps
+- The diff only contains non-test files
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/.claude-auto/validators/type-organization.md

@@ -0,0 +1,30 @@
+---
+name: type-organization
+description: Enforces inline type definitions, no standalone type files
+enabled: true
+---
+
+You are a commit validator. You MUST respond with ONLY a JSON object, no other text.
+
+Valid responses:
+{"decision":"ACK"}
+{"decision":"NACK","reason":"one sentence explanation"}
+
+**Scope:** Validate file organization for type definitions.
+
+Types must live inline where they are used, not in standalone files.
+
+**NACK if the diff adds or modifies:**
+- `types.ts` files
+- `interfaces.ts` files
+- `types/index.ts` or any file in a `types/` directory
+- Any standalone file whose sole purpose is exporting types/interfaces
+
+**ACK if:**
+- Types are defined inline in the files that use them
+- No standalone type files are created
+- The diff modifies existing files without creating type-only files
+
+**Rationale:** Types should emerge from behavior (tests). Standalone type files encourage type-first design which violates the emergent design principle.
+
+RESPOND WITH JSON ONLY - NO PROSE, NO MARKDOWN, NO EXPLANATION OUTSIDE THE JSON.

package/README.md

@@ -0,0 +1,172 @@
+# Claude Auto
+
+**Put Claude on Auto.**
+
+[![License](https://img.shields.io/badge/license-MIT-blue?style=flat-square)](LICENSE) [![TypeScript](https://img.shields.io/badge/TypeScript-Ready-blue?style=flat-square)]()
+
+---
+
+## The Problem: You're an AI Babysitter
+
+AI-assisted coding captured your cognitive load.
+
+Every session demands your full attention. Watching, nudging, correcting. You can't walk away because you don't trust the system.
+
+- You can't context-switch while supervising
+- One task at a time, full attention required
+- The bottleneck is your attention, not AI's speed
+
+That's not multiplication. That's marginally faster serial work.
+
+---
+
+## The Solution: The Quality Stack
+
+Claude Auto creates trust. Trust enables parallelization.
+
+| Component          | What It Does                                 | Result                          |
+| ------------------ | -------------------------------------------- | ------------------------------- |
+| **Auto-Planner**   | Generates plan from your requirements        | No need to specify every detail |
+| **Supervisor AI**  | Validates every commit against your criteria | Automated review                |
+| **TCR Discipline** | Test && Commit \|\| Revert                   | Bad code auto-reverts           |
+| **Auto-Continue**  | Keeps going until the plan is done           | No nudging required             |
+
+The system is trustworthy. That's what frees you.
+
+---
+
+## The Multiplier: Git Worktrees
+
+Git worktrees let you run multiple isolated workspaces from the same repo.
+
+```bash
+# Create worktrees for parallel features
+git worktree add ../feature-auth feature/auth
+git worktree add ../feature-payments feature/payments
+git worktree add ../feature-dashboard feature/dashboard
+```
+
+Each runs a Claude Auto instance. All quality-validated.
+
+| Approach                      | Features/Week |
+| ----------------------------- | ------------- |
+| Manual coding                 | 1             |
+| AI-assisted (babysitting)     | 1-2           |
+| **Claude Auto + Worktrees**   | **5-10**      |
+
+---
+
+## Three Steps
+
+### 1. Install
+
+```bash
+npx claude-auto install
+```
+
+Feed your requirements. Claude Auto auto-generates the plan with Bottles, Bursts, and Dependencies.
+
+### 2. Release
+
+Start execution and walk away. The Supervisor validates every commit. Auto-continue keeps it going until done.
+
+### 3. Multiply
+
+Open another worktree. Start another instance. Three features. Five features. All in parallel.
+
+---
+
+## The Transformation
+
+| Before (Babysitter)              | After (Bionic)                 |
+| -------------------------------- | ------------------------------ |
+| Watching one AI session          | Directing multiple workstreams |
+| Nudging, correcting in real-time | Defining, approving, releasing |
+| Serial productivity              | Parallel productivity          |
+| Marginal gains (1.5x)            | Multiplicative gains (5-10x)   |
+| Brain captured by supervision    | Brain freed for the next thing |
+
+From Babysitter to Bionic.
+
+---
+
+## Installation
+
+```bash
+# Install Claude Auto
+npx claude-auto install
+
+# Verify installation
+npx claude-auto doctor
+
+# You're ready to become Bionic
+```
+
+After installation, claude-auto automatically:
+
+- Injects hooks that validate every commit
+- Creates reminders that inject your guidelines
+- Sets up file protection via deny-lists
+- Merges settings with smart project/local overrides
+
+---
+
+## Authority: Battle-Tested
+
+Built on foundations from Kent Beck's TCR and Extreme Programming principles. Refined through production features at Auto.
+
+The on.auto team ships 5-10 features per week. Not 1-2.
+
+**[Read the origin story →](./docs/origin-story.md)**
+
+---
+
+## CLI Reference
+
+| Command                    | Description                                         |
+| ------------------------- | --------------------------------------------------- |
+| `claude-auto install`     | Install and configure claude-auto in your project   |
+| `claude-auto init`        | Initialize configuration                            |
+| `claude-auto status`      | Show symlink status for hook scripts and reminders  |
+| `claude-auto doctor`      | Diagnose installation health                        |
+| `claude-auto repair`      | Recreate broken or missing symlinks                 |
+| `claude-auto reminders`   | List active reminders with metadata                 |
+| `claude-auto clean-logs`  | Remove old log files (use `--older-than=N` to keep N recent logs) |
+
+---
+
+## Documentation
+
+| Guide                                            | Description                     |
+| ------------------------------------------------ | ------------------------------- |
+| [Getting Started](./docs/getting-started.md)     | 5-minute transformation         |
+| [Installation](./docs/installation.md)           | Detailed installation guide     |
+| [The Quality Stack](./docs/ketchup-technique.md) | Why you can walk away           |
+| [Configuration](./docs/configuration.md)         | All configuration options       |
+| [Hooks Guide](./docs/hooks-guide.md)             | Configure your supervision      |
+| [Reminders Guide](./docs/reminders-guide.md)     | Context injection system        |
+| [Validators Guide](./docs/validators-guide.md)   | Commit validation rules         |
+| [API Reference](./docs/api-reference.md)         | Programmatic access             |
+| [Architecture](./docs/architecture.md)           | System design internals         |
+
+---
+
+## Development
+
+```bash
+git clone https://github.com/BeOnAuto/claude-auto.git
+cd claude-auto
+pnpm install
+pnpm test
+pnpm build
+```
+
+---
+
+## License
+
+MIT © 2025 BeOnAuto, Inc.
+
+See [LICENSE](LICENSE) for details.
+
+---

package/bin/cli.ts

@@ -0,0 +1,6 @@
+#!/usr/bin/env node
+
+import { createCli } from '../src/cli/cli.js';
+
+const program = createCli();
+program.parse();

package/dist/bin/cli.d.ts

@@ -0,0 +1,3 @@
+#!/usr/bin/env node
+export {};
+//# sourceMappingURL=cli.d.ts.map

package/dist/bin/cli.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"cli.d.ts","sourceRoot":"","sources":["../../bin/cli.ts"],"names":[],"mappings":""}

package/dist/bin/cli.js

@@ -0,0 +1,7 @@
+#!/usr/bin/env node
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+const cli_js_1 = require("../src/cli/cli.js");
+const program = (0, cli_js_1.createCli)();
+program.parse();
+//# sourceMappingURL=cli.js.map

package/dist/bin/cli.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"cli.js","sourceRoot":"","sources":["../../bin/cli.ts"],"names":[],"mappings":";;;AAEA,8CAA8C;AAE9C,MAAM,OAAO,GAAG,IAAA,kBAAS,GAAE,CAAC;AAC5B,OAAO,CAAC,KAAK,EAAE,CAAC"}